A recent host range expansion in Junonia coenia Hubner (Nymphalidae): oviposition preference, survival, growth, and chemical defense.
The two major, competing, adaptationist hypotheses to account for the evolution and maintenance of insect dietary specialization differ in their emphases on selective forces from lower trophic levels (i.e., "bottom-up" forces) versus those from higher trophic levels ("top-down" forces) (Strong et al. 1984; Bernays 1988; Bernays and Graham 1988). According to the bottom-up hypothesis, the most important determinants of insect host-plant range are host-plant availability, apparency, and suitability as food (Ehrlich and Raven 1964; Levins and McArthur 1969; Feeny 1976; Futuyma 1976; Rhoades and Cates 1976; Cates 1980; Futuyma and Moreno 1988). In contrast, the top-down view holds that host plant-specific rates of predation and/or parasitism control the distribution and abundance of herbivores and thereby, the evolution of diet breadth (Brower 1958; Hairston et al. 1960; Strong et al. 1984; Bernays 1988; Bernays and Graham 1988; Bernays and Cornelius 1989). Specific host-plant species could provide "enemy-free space" through a variety of mechanisms including camouflage from natural enemies, exploitation of plants' physical defenses, or recycling plant defensive compounds (Lawton and McNeill 1979; Atsatt 1981).
The abilities or inabilities of ovipositing females to discriminate among plant species and the ways in which they use the available information to make such decisions, therefore, has profound impacts on diet breadth beyond food quality (Jermy 1984; Thompson and Pellmyr 1991; Bernays and Chapman 1994). They may, for example, pass up suitable food for their larvae if host plants lack required cues or present them with deterrent but harmless chemicals; or, conversely, accept unsuitable host plants that contain oviposition stimulants (Bernays and Chapman 1978; Bernays and Graham 1988; Fox and Lalonde 1993 and references therein). As a result, the sensory abilities of ovipositing females may provide either intrinsic constraint or evolutionary opportunities depending on their evolutionary lability, the consequences of oviposition "mistakes," and the relative abundance of potential hosts (Fox and Lalonde 1993; Larsson and Ekbom 1995).
Ideally, studies of insect dietary specialization would test among these adaptive and nonadaptive alternative hypotheses by investigating the areas in which they produce divergent predictions rather than seeking to confirm any single factor. This ideal, however, has been seldom realized. A plethora of studies, for instance, have investigated the major tenets of Ehrlich and Raven's (1964) coevolutionary bottom-up hypothesis. However, despite an early appreciation by some workers (e.g. Brower 1958; Hairston et al. 1960) of the potential impacts of natural enemies, these hypotheses have seldom been directly contrasted. Although this is partly due to the intuitive attractiveness of the coevolutionary hypothesis, it is also undoubtedly a consequence of the difficulties involved in identifying suitable systems in which the alternative hypotheses can be tested simultaneously.
Minimally, a system amenable to this type of study must provide: (1) quantifiable measures of female host plant oviposition preference among host plant species, host-plant food quality for larvae, and resistance to predation; (2) sufficient flexibility in the insects' feeding habits for experimental comparisons to be biologically meaningful; and (3) knowledge of the mechanisms by which predation might be host-plant species specific. The buckeye butterfly and its iridoid glycoside-containing host plants are an excellent system in which to address these questions.
Buckeyes are common inhabitants of most of coastal North America. Females oviposit and larvae feed on plants in four families, all of which contain terpene-derived iridoid glycosides (Bowers 1984). Larvae sequester two of these compounds, aucubin and catalpol, in high concentrations (Bowers and Puttick 1986; Bowers and Collinge 1992), and these compounds function as defenses against predators (Stamp 1992; De la Fuente et al. 1994/1995; Camara, in press). Pupae, however have much lower iridoid glycoside concentrations, and adults have none (Bowers and Puttick 1986; Bowers and Collinge 1992).
In the Central Valley of Northern California where I studied them, buckeyes feed primarily on a common, introduced weed, Plantago lanceolata L. (Plantaginaceae) (Scott 1972; Shapiro 1978). Plantago occurs in moist, disturbed sites such as roadsides and the margins of irrigated agricultural and athletic fields (pers. obs.). Recently, however, buckeyes have incorporated a new host plant, Kickxia elatine (L.) Dumort (Scrophulariaceae), into their diet (Shapiro 1978). Plantago is a perennial that grows as a rosette of elongate leaves with upright inflorescences projecting from the center of the rosette. Kickxia, in contrast, is an annual and its growth form is more matlike. A single plant projects many stems with a large number of small leaves and flowers over an area as large as a square meter or more. Plantago and Kickxia also differ in their iridoid glycoside profiles. Plantago leaves contain about three times the total sequesterable iridoid glycosides as Kickxia, and aucubin is approximately twice as abundant as catalpol, whereas Kickxia contains only catalpol (Camara, unpubl. data). These two host-plant species only very rarely coexist in this area (pers. obs).
Both the recency of this host shift and the availability of background information on this system make for an ideal opportunity to study the evolution of diet breadth in progress. Because sequestered iridoid glycosides function to deter predation (Stamp 1992; De la Fuente et al. 1994/1995; Camara, in press), the underlying biology of J. coenia makes it possible to contrast evolutionary responses to predation and to host-plant food quality. Growth performance and predator defense are straightforwardly quantifiable using gravimetric techniques and analytical chemical methods, respectively. In addition, because chemical defense is restricted to the larval stages in this species, it is not necessary to determine the fates of the highly mobile adults to evaluate host-plant influences on predation rates. Although adults are undoubtedly subject to predation, it is hard to imagine how their probability of being eaten could be host-plant specific after they pupate.
Similar plant-insect systems of nymphalid butterflies that feed on iridoid-glycoside containing host plants have been used by other investigators to examine host plant-specific population differentiation, but no generalizations have yet emerged from these efforts. For example, Bowers (1986) studied two populations of Euphydryas chalcedona that feed on two different Penstemon species (Scrophulariaceae). She found that prediapause larvae from one of the populations showed local adaptation to their available host-plant species, but that another was more flexible. Both Rausher (1982) and Thomas et al. (1987) studied populations of Euphydryas editha, but they obtained different results. In Rausher's study of two populations, each population grew better on its own host plant, but Thomas and coworkers found no evidence of such local adaptation when they compared a population that had recently expanded its host range to include P. lanceolata to others that had not. However, Thomas et al. (1987) and several other studies have found population-level differences in oviposition behavior in this species (Singer 1971; Singer et al. 1989). Furthermore, Singer et al. (1988) demonstrated in a single population that oviposition preferences in E. editha are heritable and correlated with larval growth.
These sometimes conflicting results are difficult to explain considering the similarities of the basic biologies of these closely related species. None of these studies, however, addressed any top-down influences on diet breadth. If this neglected factor is an important influence on the host range of these species, investigations that incorporate it may provide more generalizable results. To date, only one investigation of this type included traits related to predation: Bowers et al. (1992) found that E. phaeton populations with different host plants available to them survived better on their own host-plant species, but that the two populations showed similar oviposition and larval feeding preferences. Both populations preferred the host plant richer in sequesterable iridoid glycosides, suggesting that ovipositing females and foraging larvae may prefer high levels of chemical defense over nutritional quality.
In this paper, I examine the oviposition preferences, growth performance, and chemical defense of buckeyes found in sites that offer only Kickxia or Plantago as hosts. I reason that if bottom-up factors normally preclude host shifts, then local populations that have adopted a novel host plant should demonstrate enhanced performance on the only host plant available to them. On the other hand, if top-down selection maintains host-plant specificity, then populations should be locally adapted for more effective chemical defense.
In either case, one might also expect a relationship between female oviposition preference and the dominant selective pressure on their behavior. These correlations have been a major working hypothesis in the evolution of specialized feeding habits in response to bottom-up selection, but most studies have focused on the population level. As emphasized by Via (1986), however, it is the correlations at the individual level, not the correlations among population means, that are relevant to evolutionary questions. Positive associations, Although not necessary for local adaptation, promote rapid divergence by facilitating coordinated evolution of preference and traits under selection on each host plant (reviewed in Via 1990; Thompson and Pellmyr 1991). Only one study, however, has demonstrated positive oviposition/larval performance correlations at the level of individuals (Singer, et al. 1988).
If, on the other hand, females show no preference for their local host-plant species and/or there are negative relationships between their oviposition preferences and either food quality or chemical defense, we might conclude that diet breadth is not shaped by adaptation to either top-down or bottom-up selection. Rather, their oviposition preferences could be viewed as an intrinsic property of the sensory system of the animals themselves, perhaps an evolutionarily maladaptive "dead-end," if the species is trapped on inferior or rare host-plant species by behavioral inflexibility (Moran 1988).
I also examined the underlying genetic potential for population divergence in physiological adaptation to host plants as food and as sources of chemical defense, trade-offs among host-plant species in both performance and chemical defense, and trade-offs between growth and chemical defense within host-plant species. A lack of heritable variation for any trait would prevent an evolutionary response despite even strong selective pressure on phenotypes, a distinction emphasized by Haldane (1954). Across-diet trade-offs promote local specialization and even sympatric divergence of host races because individuals that choose host plants inappropriately suffer decreased growth. Even without strict trade-offs, however, the divergence of populations with different host plants available is more likely if the genetic architecture precludes the evolution of "super genotypes" which perform better under all circumstances (reviewed by Via 1990). That is, if "a jack of all trades is a master of none," then populations will evolve to local optima. Within-diet trade-offs examine the potential costs of chemical defense. It is, of course, the balance of such costs versus the benefits that determine whether or not any particular host-plant species will be included in or excluded from the diet.
MATERIALS AND METHODS
I conducted this experiment between 15 June and 10 September 1993. I first collected adult, female buckeye butterflies from two naturally occurring populations in the Sacramento area of California's Central Valley. The Discovery Park site (DP, Sacramento County) is an intermittently mown field in which the only available host plant for buckeye larvae is P. lanceolata. Adults and larvae feeding on Plantago were extremely common at this site. The West Sacramento site (WS, Yolo County) is an area between railroad tracks and agricultural fields where K. elatine is abundant. Both adults and larvae were less abundant here than at DP. Another reported host plant, Lippia nodiflora Michx. (Verbenaceae) (Shapiro 1974), was present at the WS site, but was much less common than K. elatine, and P. lanceolata was absent. Adult butterflies at this site were much more common in the vicinity of Kickxia than Lippia; and although some adults were observed to nectar on Lippia, I never observed any females ovipositing on this plant during the course of this study. In preliminary rearings in the laboratory, buckeye larvae fed Lippia grew extremely slowly and most died in the first or second instar (Camara, pers. obs.). I thoroughly searched these Lippia plants several times during the course of this study, but never found larvae on them. In contrast, I commonly observed females ovipositing on Kickxia, and found naturally occurring larvae on these plants. The two study sites are about nine kilometers apart.
TABLE 1. Dates and sources of females for oviposition preference tests. (DP = Discovery Park, WS = West Sacramento.) Date tested Population Number of females tested 17 June DP 10 22 June WS 8 24 June DP 8 8 July WS 10 13 July WS 10 16 July DP 10 18 July WS 10 17 July WS 10
All females were transported to the University of California at Davis, held for 48 h in large wood and wire mesh cages (1 m x 1 m x 1 m) in a greenhouse with natural photoperiod, and provided with an abundance of a dilute honey/water solution on cotton balls in 1-oz. plastic cups. No host plants were provided during this time. This holding period was intended to serve two functions. First, by ensuring that all females had not encountered host plants for at least two days, it minimized any carry-over effects of previous oviposition experience. Second, it standardized the influence of nutritional state and egg load on oviposition tests.
After the holding period I tested, females for oviposition preferences between Plantago and Kickxia (Table 1) by placing them individually for 24 h into cylindrical aluminum window screen cages (0.5 m diameter x 0.5 m height) in which two similarly sized sprigs each of Kickxia and Plantago were held in aquapics. The aquapics were fastened to the walls of the cages by wire loops such that the leaves leaned toward the center of the cage, but did not touch each other, and a small bouquet of freshly cut flowers from ornamental plantings on the campus of the University of California at Davis was placed in the center of each cage as a nonhost perch. I placed females onto these flowers at the start of each trial to avoid biasing their initial encounters with the host plants. The greenhouse in which I did these tests was heavily whitewashed, and although light levels were not measured, the environment was uniformly bright and the light was diffuse rather than direct. At the end of the trial, I removed the females and counted all eggs on each plant sprig. Data from females that produced fewer than 15 eggs during the 24-h test were not included in analyses of oviposition preference, nor were those females used in the remainder of this experiment because it was considered unlikely that they would produce the required number of offspring for the next phase of the experiment (see below).
Those females that produced 15 or more eggs during the oviposition tests were then placed in individual plastic boxes (12.5 x 17.5 x 7 cm) in a growth chamber that maintained a constant temperature of 27 [degrees] C and a consistent photoperiod (15:9 L:D). Each female was provided with Plantago leaves in aquapics as an oviposition substrate. To maximize the number of females that did produce sufficient offspring, they were hand-fed a dilute honey solution daily and provided with cotton balls moistened with the same dilute honey solution between feedings. Even so, not all females produced the 40 fertile eggs required to complete this part of the experiment. In total, 19 females from the WS population and 23 from DP were used in this part of the experiment, with a total of 1680 neonate larvae.
I removed all eggs from the Plantago leaves to small petri dishes every few days, before any had hatched, and the leaves in the aquapics were replaced with fresh ones as necessary. The eggs were checked several times daily for hatching. As they became available, 40 neonate offspring from each female were moved to individual, 2-oz. plastic cups sealed with both moistened cardboard lids to humidify the air in the cups and plastic lids to maximize the retention of this humidity. These precautions minimized the desiccation of the food leaves fed to the larvae. Of the 40 larvae from each female, 20 were fed freshly collected leaves of Kickxia and 20 were fed Plantago, and all larvae were reared under the same growth chamber conditions described above. I inspected all cups daily, and recorded any deaths. If necessary, I provided fresh leaves at these times, and replaced all food every third day regardless of how much had been consumed or added. Within each group of 20, I randomly chose 10 larvae from the outset to be harvested at the fourth larval molt, because iridoid glycoside assays require destructive sampling (see below). I designated the other 10 to be reared to the pupal stage. As the larvae approached their designated developmental stage, they were checked four times daily, and immediately after they had completed either the fourth larval molt or the pupal molt, they were frozen in liquid nitrogen and freeze-dried. Freeze-dried larvae and pupae were stored frozen at - 70 [degrees] C until the end of the experiment, when they were transported to Boulder, Colorado, and stored in a commercial freezer until they could be weighed and analyzed.
Frozen larvae and pupae were removed from the freezer in small batches and stored overnight under vacuum to ensure that they were completely desiccated. Each larva and pupa was then weighed to the nearest 0.01 mg using a Mettler HK-60 analytical balance.
I quantified the levels of iridoid glycosides sequestered by fifth-instar larvae using gas liquid chromatography (Gardner and Stermitz 1988). Briefly, I extracted the finely ground larvae in 5 mL of methanol, and after filtering the extracts, rinsing the solids with another 9 mL of methanol, and drying the filtrate, removed nonpolar compounds by partitioning the extracts into water-soluble and ether-soluble fractions. I then trimethylilylated a 10% aliquot of the aqueous fraction in 0.1 mL of TRI-SYL Z (Pierce Chemical Company) and injected 1 [[micro]liter] into a Hewlett Packard 5890 gas chromatograph equipped with a split/splitless injector, FID detector, and a 3392A Hewlett Packard integrator. I used a capillary column (J&W Scientific DB-1, length = 30 m, I.D. = 0.32 mm, film thickness = 0.1 [[micro]meter]). The temperature and flow program used was Gardner and Stermitz's (1988) "Program A."
To test for population differences in oviposition preference, I grouped the oviposition preferences of individual females into categories based on how they distributed their eggs among host-plant species. Females that concentrated 75% or more of their eggs on one host-plant species were classified as "preferring" that species. I classified females that spread their ovipositions more evenly as "neutral." These divisions reflect the strong bimodality of the overall distribution of females' preferences [ILLUSTRATION FOR FIGURE 1 OMITTED]. Furthermore, they do not result from repeated oviposition on the first plant sprig encountered. Nearly all females spread their eggs among sprigs within a host-plant species even when they showed strong preferences. In addition, this classification scheme is highly conservative in that females were required to show very strong biases before they were categorized as preferring either host-plant species. I used logistic regression to test for source population differences in female oviposition preference using the SAS CATMOD procedure (SAS Institute 1987).
I also used logistic regression to test for effects of source population, host-plant diet and their interaction on the survivorship of larvae. Because larvae were designated from the outset of the experiment to be frozen as either newly molted fifth-instar larvae or pupae, I performed separate logistic regressions for survival to each stage.
I analyzed continuously distributed measures of performance (weights and development times) using separate multivariate analyses of variance (MANOVA) on the larval and pupal responses. In each of these analyses, the dry weight and development time of each pupa or larva were analyzed simultaneously. Both were log transformed prior to these analyses to meet the assumptions of normality and homogeneity of variances, and SAS Type III sums of squares were used, because mortality made the experimental design slightly unbalanced. The linear model for the multivariate analyses used both fixed and random factors. These factors and the error terms used to test each are listed in Table 2. Note that there are two hypothesis tests for dam effects that use different error terms (see below for their different interpretations). I refer to these effects as "dam-over-interaction" and "dam-over-error."
TABLE 2. Factors tested in multivariate and univariate analyses of variance. Source Type Error term Host plant Fixed Host plant*dam (population) Population Fixed Dam(population) Host plant*population Fixed Host plant*dam (population) Dam(population)/ Random Host plant*dam interaction (population) Dam(population)/error Random Error Host plant*dam Random Error (population)
The main effects of host plant in these analyses test the null hypothesis that the two host-plant species differ in overall food quality for both populations. Main effects of population test for differences in growth performance between the populations, regardless of the host-plant species. Host plant x population interactions test the null hypothesis that the larvae from the two populations respond in the same way to the host-plant treatment. That is, local populations are locally adapted if such interactions are significant and of the crossing type (Via 1984, 1990). Dam-over-interaction effects provide a test of whether or not there are significant tradeoffs among host plants (Fry 1992). These tests are two-tailed and reject the null hypothesis that the correlation of family-means among host plants is zero in favor of the alternative that the true correlation is positive if P [less than] 0.025 and negative if P [greater than] 0.975. Negative correlations of family-means across hosts would indicate trade-offs, such that genotypes that perform better on one host predictably do poorly on others. In contrast, dam-over-error effects test for significant heterogeneity among family-means and, therefore broad-sense genetic variation when both host plants are considered simultaneously. Finally, host plant X dam interactions test for structuring of genetic variation, such that there is broad-sense genetic variation for specialization on different host-plant species (Via 1984; Falconer 1989). That is, even without strict trade-offs, genotypes change ranks on different host-plant species.
I also did univariate ANOVAs to dissect out the contribution of each scalar in the response vector in the MANOVA analyses. Hypothesis tests use the same linear model as the multivariate analyses and, again, Type III sums of squares. The RANDOM option in the SAS GLM procedure was used to produce appropriate hypothesis tests of the mixed-model effects, but the dam-over-error tests were hand calculated because SAS does not normally produce these tests for a random factor in a mixed-model design. Due to the slight imbalance in the design, error terms in the SAS tests were, in most analyses, fractional combinations of several mean squares. Table 2 still, however, represents the majority contribution to each error term. The hand-calculated dam-over-error tests were not affected by this complication.
I analyzed iridoid glycoside concentrations in the same way as performance measures. Joint analyses of the angular-transformed proportions of larval dry weight consisting of aucubin and catalpol were carried out using MANOVA, and significant multivariate effects were followed up with univariate analysis of variance.
Finally, I tested for relationships between the oviposition preference of individual females and the mean performance and chemical defense measures of their offspring on each host-plant species. For these tests, I used a continuous measure of female oviposition preference calculated as the proportion of each females' eggs laid on Plantago in the oviposition trial. This index ranges from zero to one, with values above 0.5 indicating preference for Plantago and values below 0.5 indicating preference for Kickxia. I performed separate Spearman rank correlation analyses of family-means for each host plant species to test the hypothesis that there are trade-offs between growth and chemical defense within host plants. Significant negative family-means correlations between chemical defense and weights would be evidence of such trade-offs, as would positive correlations between chemical defense and development time. Lack of correlation or correlations of opposite sign would indicate an absence of such trade-offs.
Oviposition Preferences and Survivorship
The two populations studied differed significantly in how females distribute their eggs among the two host-plant species [ILLUSTRATION FOR FIGURE 2 OMITTED]. A majority of the females from each population preferred the predominant host-plant species in the area from which I collected them. Survival was high, with over 70% of the larvae surviving in all host-plant, source population combinations [ILLUSTRATION FOR FIGURE 3 OMITTED]. Survivorship to both the fifth-instar (Table 3a) and pupal stage (Table 3b) were significantly affected by the host-plant species fed to the larvae, with significantly higher survival on Plantago regardless of the population from which females were collected [ILLUSTRATION FOR FIGURE 3 OMITTED]. There was no significant effect of either source population or of host plant x population interaction on survival to the fifth-instar or the pupal stage (Table 3).
Effects on larval and pupal performance characters are complex. All main and interaction effects are significant in the MANOVA analysis of newly molted fifth-instar weight and development time from egg to the fifth instar (Table 4a), and all factors except host plant x population interaction are significant for pupal weight and development time from egg to pupa (Table 4b).
Turning to the univariate analyses, larval and pupal weights respond differently to the host-plant treatments. Fifth-instar larvae are heavier on Kickxia than on Plantago, but the reverse is true for pupae [ILLUSTRATION FOR FIGURE 4 OMITTED]. Both effects are statistically significant (Table 5a,b). Both fifth-instar larvae and pupae from WS mothers were significantly larger than those from DP regardless of the host plant on which they were raised ([ILLUSTRATION FOR FIGURE 4 OMITTED], Table 5a,b). Neither the univariate host plant x population interactions for the weight of larvae, nor those for the weight of pupae, are significant (Table 5a,b). Both types of univariate dam effects on dry weight are significant for larvae and for pupae, but host plant x dam interaction effects on weight measures are significant only for fifth instars ([ILLUSTRATION FOR FIGURE 5 OMITTED]; Table 5a,b).
In contrast, development time shows consistent significant main effects of host plant for both pupae and larvae ([ILLUSTRATION FOR FIGURE 6 OMITTED]; Table 5c,d). Development on Kickxia is significantly longer than on Plantago for both fifth instars and pupae. Source population effects on development time are not significant ([ILLUSTRATION FOR FIGURE 6 OMITTED]; Table 5c,d), but host plant X population interactions are significant for both larvae and pupae ([ILLUSTRATION FOR FIGURE 7 OMITTED]; Table 5c,d). Interestingly, however, both populations perform relatively better on the other population's host plant: DP larvae reach both the fifth instar and pupation sooner on Kickxia than do WS larvae and WS larvae develop faster on Plantago than DP caterpillars. Dam-over-interaction effects on development time are not significant for larvae or pupae, but dam-over-error effects are significant for both larval and pupal development time. (Table 5c,d). Host plant x dam interaction effects on development time are significant for both larvae and pupae ([ILLUSTRATION FOR FIGURE 7 OMITTED]; Table 5c,d).
MANOVA analyses of allelochemical sequestration show significant multivariate effects of all factors in the model (Table 6). Univariate main effects of host plant species on chemical defense (Table 7; [ILLUSTRATION FOR FIGURE 8 OMITTED]) show that larvae from both source populations sequester significantly more of both iridoid glycosides when fed Plantago, reflecting their greater availability in this plant. Although significant for both iridoid glycosides, population effects differ in sign depending on which iridoid glycoside is being examined ([ILLUSTRATION FOR FIGURE 8 OMITTED]; Table 7a,b). DP larvae sequester more aucubin when it is available in their diet [ILLUSTRATION FOR FIGURE 8A OMITTED], but WS larvae sequester more catalpol regardless of host plant [ILLUSTRATION FOR FIGURE 8B OMITTED]. Host plant x population effects are significant only for aucubin (Table 7a), but this is largely artifactual because aucubin is virtually unavailable in Kickxia [ILLUSTRATION FOR FIGURE 9A OMITTED]. Host plant x dam interaction is present for both iridoid glycosides ([ILLUSTRATION FOR FIGURE 9 OMITTED]; Table 7a,b).
Preference/Performance Correlations and Costs of Chemical Defense
Spearman rank correlations between female oviposition preference and measures of either growth performance or chemical defense of their offspring were not significant for any response variable on either host (all P [greater than] 0.05). This was equally true when both populations were combined into a single analysis or separate analyses were performed for each population.
TABLE 3. Maximum-likelihood analysis of variance on (A) fifth-instar survival; and (B) pupal survival. Source df [[Chi].sup.2] p A. Intercept 1 283.62 [less than] 0.0001 Host plant 1 27.91 [less than] 0.0001 Population 1 0.53 0.468 Host plant*population 1 1.58 0.209 B. Intercept 1 257.18 [less than] 0.0001 Host plant 1 25.16 [less than] 0.0001 Population 1 1.65 0.1991 Host plant*population 1 0.02 0.8843
[TABULAR DATA FOR TABLE 4 OMITTED]
Separate correlation analyses of the family-means for chemical defense measures with the family-means of performance measures within each population reveal a strong pattern. DP larvae show positive correlations between catalpol concentration and both weight measures and development times when these larvae are fed Kickxia (Table 8a). On a Plantago diet, there are no such correlations (Table 8b): the correlation between catalpol content and fifth-instar development time becomes negative; and aucubin concentration is correlated with extended development times for pupae. The WS population, in contrast, shows a negative correlation between pupal weight and catalpol concentration when fed Kickxia (Table 8c), and a positive correlation between aucubin levels and development time when fed Plantago (Table 8d).
Females collected from these two populations show marked preferences for the host-plant species available at their collection site [ILLUSTRATION FOR FIGURE 2 OMITTED], but this result should be interpreted with caution, as the females were collected from the field and are likely to have prior oviposition experience. I found very little evidence that larvae from these populations are physiologically adapted to better utilize the plants available to them as food. Larvae from both populations survive better on P. lanceolata than on K. elatine, but there are no significant differences in survival between the two populations and no significant host plant x population interactions for survivorship [ILLUSTRATION FOR FIGURE 3 OMITTED]. Larvae from the WS population develop into larger fifth-instar larvae and pupae than do larvae from the DP population regardless of which host plant they are fed, and host-plant effects on weight switch sign between the fifth instar and pupation [ILLUSTRATION FOR FIGURE 4 OMITTED]. This is a curious pattern that is probably due to selective mortality of larger larvae during the fifth instar, though I have no direct evidence of this. In terms of weight at both the fifth-instar and pupal stages, larvae from WS are superior performers on both host-plant species.
There is evidence, however, of host-plant-specific differentiation in development time as evidenced by significant host plant x population interactions for both larval and pupal development times ([ILLUSTRATION FOR FIGURE 6 OMITTED]; Table 5c,d). But, examination of these interactions reveals apparent maladaptiveness. Larvae from both populations develop more quickly on the other population's host plant. For development times then, the data reveal a pattern of local maladaptation, which is difficult to interpret if bottom-up selection were important in the evolution of diet breadth in this species.
In contrast, the patterns of local differentiation in chemical defense are more amenable to the interpretation that local populations have adapted to the host plant available to them. Larvae from neither population sequester detectable levels of aucubin when fed Kickxia, but larvae from the DP population sequester more aucubin than do larvae from the WS [TABULAR DATA FOR TABLE 5 OMITTED] population when they feed on diets that contain aucubin [ILLUSTRATION FOR FIGURE 9A OMITTED]. As well, larvae from the WS population sequester more catalpol than larvae from the DP population regardless of which host-plant species they are fed [ILLUSTRATION FOR FIGURE 9B OMITTED]. Thus, WS larvae are more effective at sequestering the only iridoid glycoside compound available in their normal host-plant species (catalpol), and the DP population is more effective in sequestering the iridoid glycoside compound that is available only to them (aucubin). The WS larvae are catalpol specialists and show reduced ability to sequester aucubin.
I interpret this combination of results - evidence of oviposition specialization, lack of evidence for local adaptation in performance, and evidence of sequestrative specialization - as indirect evidence supporting the hypothesis that natural enemies are an important determinant of diet breadth in J. coenia. That some females accept Kickxia for oviposition is a prerequisite for such a host range expansion to occur, and would put some neonate larvae in the position of having to survive on Kickxia or die in the WS population. If the barriers to their survival were related to food quality, I would expect that those larvae that do survive after being forced by their mothers to feed on Kickxia would show enhanced abilities to grow and develop on this foodplant. This prediction is not supported by my data. If, however, the barriers to survival on Kickxia are related to the avoidance of natural [TABULAR DATA FOR TABLE 6 OMITTED] enemies, I would predict that those larvae would develop better mechanisms to defend themselves - in this case, increased ability to sequester the iridoid glycosides available to them. This prediction is supported by my data.
TABLE 7. Analysis of variance on (A) aucubin concentration; and (B) catalpol concentration. Type Source df III MS F P A. Host plant 1 0.523 207.436 0.0001 Population 1 0.016 6.538 0.0142 Host plant*population 1 0.016 6.491 0.0145 Dam(population)/ interaction 40 0.003 1.003 0.4956 Dam(population)/error 5.918 [less than] 0.0001 Host plant*dam (population) 40 0.003 5.979 [less than] 0.0001 Error 612 0.0005 B. Host plant 1 0.134 108.892 [less than] 0.0001 Population 1 0.028 9.736 0.0032 Host plant*population 1 0.002 1.536 0.2213 Dam(population)/ interaction 40 0.003 2.477 0.0025 Dam(population)/error 5.484 [less than] 0.0001 Host plant*dam (population) 40 0.001 2.190 0.0001 Error 612 0.0006
The genetic architecture underlying these population-level effects is complex. The significant multivariate and univariate dam-over-interaction effects for both larval and pupal growth performance (Tables 4, 5) and chemical defense (Table 6) indicate that the correlations of family-means performance characters among diets are positive, so there is no evidence of across-diet trade-offs. In addition, multivariate evolutionary responses to top-down and bottom-up selection are not likely to be constrained by a lack of genetic variation, because there is considerable heterogeneity among families in performance measures as indicated by significant dam-over-error effects for both chemical and performance measures (Tables 4, 6).
Significant dam-over-error and host plant X dam interactions for fifth-instar weight, pupal weight, and pupal development time indicate that there is genetic potential for divergence of host-plant-specific populations in performance, but there is no evidence that this potential has been realized. One interpretation of this pattern is that there has been no selection pressure to promote this divergence. That is, bottom-up selection in relation to the inclusion of Kickxia as a food plant has been weak at best.
Univariate dam-over-interaction effects on chemistry are significant only for catalpol, but the lack of significance for aucubin is probably an artifact of Kickxia containing virtually none of this iridoid glycoside. This would both reduce the component of variation in aucubin content attributable to families (all families have none when fed Kickxia), and increase the component of variation due to host plant x dam interaction. Both of these influences would tend to reduce the F-statistics for these tests. There is, however, no evidence of negative correlation in chemical defense across hosts. Evolutionary changes in chemical defense, however, are not likely to be constrained by a lack of available variability because there is substantial heterogeneity among families as estimated by dam-over-error effects (Tables 6, 7).
Additionally, both iridoid show significant host plant x dam interaction (Table 7), which would facilitate the evolution of sequestrational specialists on the two host-plant species because they indicate that families that sequester more iridoid glycosides on one host than others do not necessarily do so on other host plants. Different genotypes would be favored by selection on chemical defense in each environment. In contrast to the performance results, however, this genetic propensity for divergence seems to have been selected upon. The two populations have different abilities to sequester iridoid glycosides.
The lack of significant correlations between female oviposition preference and either larval performance or chemical defense indicates that these traits evolve independently. The observed population divergence in sequestrative ability is, therefore, probably a result of strong local selection from natural enemies because it is not facilitated by these correlations.
The two populations also differ in the costs they pay for the acquisition of chemical defenses. Both pay higher costs on their normal host-plant species (Table 8). The DP population, for example, shows significant positive correlations between fifth-instar and pupal weights and catalpol concentration when fed Kickxia, accompanied by negative correlations with development times to both stages. When fed Plantago, however, this population shows similar negative correlations with development time, but the positive correlations with weight measures is absent. WS larvae show a negative correlation between catalpol concentration and pupal weight [TABULAR DATA FOR TABLE 8 OMITTED] only when fed Kickxia, their normal host plant. When WS larvae eat Plantago, increased aucubin concentration is significantly correlated with longer development times, but this population is not exposed to aucubin in its normal habitat. While these patterns must be interpreted with caution in the absence of knowledge of other biochemical differences between the host-plant species, they support the idea that these populations are under selection to maximize iridoid glycoside concentration despite these costs. Each population achieves a higher concentration of the iridoid glycoside compound most available in its natural host plant than does the other population, even though these higher concentrations are detrimental to growth.
These data also help to explain the disparity of results obtained in other studies of local adaptation by nymphalid butterflies to iridoid glycoside-containing host plants (see introduction). Most studies have not addressed the role of predators, and the one that includes data on chemical defense (Bowers et al. 1992) only points out that different host-plant species provide different levels of chemical defense. If my results can be generalized, these studies could represent a continuum in that they may simply be different stages in the adoption of a novel host-plant species. This study examines a very early stage in that process, when adaptation to natural enemies is ongoing, and physiological adaptation to food plants may be minimal. Over time, however, if predators maintain this association, one would expect that a reduction of mortality due to predation or a decrease in the genetic variation for chemical defense would eventually put a premium on physiological adaptation and efficient assimilation of host-plant tissue.
In a broader context, this study has several important implications for the study of insect host range and ecological specialization in general. The first of these is somewhat obvious. The current challenge is to design studies that test between alternative biological hypotheses explaining the evolution of host range, rather than confirm single-factor theories (Janzen 1988). Although this study suggests that predators may be driving local adaptation in this species, it could also have detected effects of host-plant chemistry. In part, the historical overemphasis on bottom-up selection on insect diet breadth (Strong et al. 1984; Bernays and Graham 1988) is the result of a lack of such experiments.
For instance, one of the most widely accepted examples of coevolutionary determination of host range in phytophagous insects is the association between plants containing coumarins and their specialist herbivores (recently reviewed in Berenbaum 1990). These relationships do show remarkable patterns in phylogenetic relationships, diversity of herbivores feeding on particular plant species with more or less toxic coumarins, and specific mechanisms in insects to detoxify these compounds, all of which are consistent with Ehrlich and Raven's (1964) coevolutionary scenario. However, all of these patterns are also consistent with the hypothesis that top-down selection maintains these associations. Hay and coworkers (Hay et al. 1990; Duffy and Hay 1994), for example, have recently demonstrated in marine systems that macroherbivores feeding on seaweeds consume large numbers of phytophagous amphipods. They argue that bad foods for these macroherbivores make good houses for the amphipods because they avoid casual predation. These amphipods are biochemically adapted to chemically defended seaweeds, but probably not because they have coevolved. Could this scenario also apply to coumarin-containing plants and their herbivores? Could avoiding large grazers, now virtually eliminated from most of our terrestrial ecosystems, be the reason that some insects specialize on highly toxic plants? Unfortunately, we cannot answer the question because there are no studies of host-plant-specific predation rates in these organisms.
This is not to deny the importance of nutritional adaptation to host plants and their various defenses. Dethier (1954) and Futuyma (1983), for example, have argued that biochemical adaptation to new host-plant species occurs after behavioral specialization This study found no support that bottom-up selective forces are important in the initial stages of this host range expansion, and thus suggests that buckeyes initially evolve responses to predators rather than to plant defenses. Natural enemies, therefore, may initially control host range and thereby the intimacy of the relationships between plants and insects that feed on them (Jermy 1984). Under this scenario, a new host-plant species may be first incorporated into an insect's diet due to initial oviposition "mistakes," after which specialization imposed by predators may result in subsequent coevolution. Specialists would then be under selection for efficient acquisition of nutrients from protected plant tissue, and plants may respond. Coevolution, then, cannot be said to drive the specialized feeding habit, but rather to result from the imposition of specialization by natural enemies.
The second major implication of this study is that investigations of the preconditions necessary for the divergence of host-plant-specific populations or host races do not necessarily imply that those responses have or will occur. Although the populations studied in this investigation show the genetic potential for divergence in both growth performance and chemical defense (significant host plant X dam interactions), only the chemical traits have realized this potential.
This research was funded by National Science Foundation Dissertation Improvement Grant # DEB-9212542 to the author, a Dean's Small Grant from the University of Colorado, and a Grant-in-Aid from the Van Riper fund of the University of Colorado Museum. K. Darrow assisted with the collection and care of the insects. A. Sudborough and J. Kleffner helped to sort the samples and to prepare them for gas chromatography. A. Shapiro and the University of California at Davis provided laboratory and growth chamber space, and M. D. Bowers allowed me to use her gas chromatograph and laboratory facilities. M.D. Bowers, S. Susnowitz, Y. Linhart, and P. Diggle commented on earlier drafts of the manuscript.
ATSATT, P. R. 1981. Lycaenid butterflies and ants: selection for enemy-free space. Am. Nat. 118:638-654.
BERENBAUM, M. R. 1990. Evolution of specialization in insect-umbellifer associations. Annu. Rev. Entomol. 35:319-343.
BERNAYS, E. A. 1988. Host specificity in phytophagous insects: selection pressure from generalist predators. Entomol. Exp. Appl. 49:131-140.
BERNAYS, E. A., AND R. F. CHAPMAN. 1994. Host-plant selection by phytophagous insects. Chapman and Hall, New York.
BERNAYS, E. A., AND R. G. CHAPMAN. 1978. Plant chemistry and acridoid feeding behavior. Pp. 100-141 in J. B. Harborne, ed. Coevolution of plants and animals. Academic Press, New York.
BERNAYS, E. A., AND M. L. CORNELIUS. 1989. Generalist caterpillar prey are more palatable than specialists for the generalist predator Iridomyrmex humilis. Oecologia (Berlin) 79:427-430.
BERNAYS, E., AND M. GRAHAM. 1988. On the evolution of host specificity in phytophagous arthropods. Ecology 69:886-892.
BOWERS, M. D. 1983. The role of iridoid glycosides in host plant specificity of checkerspot butterflies. J. Chem. Ecol. 9:475-493.
-----. 1984. Iridoid glycosides and host-plant specificity in larvae of the buckeye butterfly, Junonia coenia (Nymphalidae). J. Chem. Ecol. 10:1567-1577.
-----. 1986. Population differences in the checkerspot butterfly, Euphydryas chalcedona. Entomol. Exp. Appl. 40:61-69.
BOWERS, M. D., AND S. K. COLLINGE. 1992. Fate of iridoid glycosides in different life stages of the buckeye, Junonia coenia, (Lepidoptera: Nymphalidac). J. Chem. Ecol. 18:817-831.
BOWERS, M. D., AND G. M. PUTTICK. 1986. The fate of ingested iridoid glycosides in lepidopteran herbivores. J. Chem. Ecol. 12: 169-178.
BOWERS, M. D., N. E. STAMP, AND S. K. COLLINGE. 1992. Early stage of host range expansion by a specialist herbivore, Euphydryas phaeton (Nymphalidae). Ecology 73:526-536.
BROWER, L. P. 1958. Bird predation and food plant specificity in closely related procryptic species. Am. Nat. 92:183-187.
CAMARA, M. D. In press. Do sequestered iridoid glycosides protect buckeye caterpillars from predators? A field test. J. Chem. Ecol.
CATES, R. G. 1980. Feeding patterns of monophagous, oligophagous, and polyphagous insect herbivores: the effect of resource abundance and plant chemistry. Oecologia (Berlin) 16:22-31.
DE LA FUENTE, M., L. A. DYER, AND M. D. BOWERS. 1994/1995. The iridoid glycoside, catalpol, as a deterrent to the predator Camponotus floridanus (Formicidac). Chemoecology 5/6:13-18.
DETHIER, V. G. 1954. Evolution of feeding preferences in phytophagous insects. Evolution 8:33-54.
DUFFY, J. E., AND M. E. HAY. 1994. Herbivore resistance to seaweed chemical defense: the roles of mobility and predation risk. Ecology 75:1291-1306.
EHRLICH, P. R., AND P. H. RAVEN 1964. Butterflies and plants: a study in coevolution. Evolution 18:586-608.
FALCONER, D. S. 1989. Introduction to quantitative genetics. Wiley, New York.
FEENY, P. 1976. Plant apparency and chemical defense. Recent Adv. Phytochem. 10:1-40.
FOX, C. W., AND R. O. LALONDE. 1993. Host confusion and the evolution of insect diet breadths. Oikos 67:577-581.
FRY, J. D. 1992. The mixed-model analysis of variance applied to quantitative genetics: biological meaning of the parameters. Evolution 46:540-550.
FUTUYMA, D. J. 1976. Food plant specialization and environmental predictability in Lepidoptera. American Naturalist 110:285-292
-----. 1983. Evolutionary interactions among herbivorous insects and plants. Pp. 207-231 in D. J. Futuyma and M. Slatkin, ed. Coevolution. Sinauer, Sunderland, MA.
FUTUYMA, D. J., AND G. MORENO. 1988. The evolution of ecological specialization. Annu. Rev. Ecol. Syst. 19:207-233.
GARDNER, D. R., AND F. R. STERMITZ. 1988. Host plant utilization and iridoid glycoside sequestration by Euphydryas anicia (Lepidoptera: Nymphalidae). J. Chem. Ecol. 14:2147-2168.
HAIRSTON, N. G., F. E. SMITH, AND L. B. SLOBODKIN. 1960. Community structure, population control, and competition. Am. Nat. 94:421-425.
HALDANE, J. B. S. 1954. The measurement of natural selection. Proc. 9th Congr. Genet. 1:480-487.
HAY, M. E., J. E. DUFFY, AND W. FENICAL. 1990. Host-plant specialization decreases predation on a marine amphipod: an herbivore in plant's clothing. Ecology 71:733-743
JANZEN, D. H. 1988. On the broadening of insect-plant research. Ecology 69:905.
JERMY, T. 1984. Evolution of insect/host plant relationships. Am. Nat. 124:609-630.
LARSSON, S., AND B. EKBOM. 1995. Oviposition mistakes in herbivorous insects: confusion or a step towards a new host plant? Oikos 72:155-160.
LAWTON, J. H., AND S. McNEILL. 1979. Between the devil and the deep blue sea: on the problem of being a herbivore. Syrup. Brit. Ecol. Soc. 20:223-244.
LEVINS, R., AND R. H. McARTHUR. 1969. An hypothesis to explain the incidence of monophagy. Ecology 50:910-911.
MORAN, N. A. 1988. The evolution of host-plant alternation in aphids: evidence for specialization as a dead-end. Am. Nat. 132: 681-706.
RAUSHER, M. D. 1982. Population differentiation in Euphydryas editha butterflies: larval adaptation to different hosts Evolution 36:581-590.
RHOADES, D. F., AND R. G. CATES. 1976. Toward a general theory of plant antiherbivore chemistry. Recent Adv. Phytochem. 10: 168-213.
SAS INSTITUTE. 1987. SAS/STAT guide for personal computers. Vers. 6 ed. SAS Institute, Inc., Cary, NC.
SCOTT, J. A. 1972. Comparative mating and dispersal systems in butterflies. Ph.D. diss. Univ. of California, Berkeley.
SHAPIRO, A. M. 1974. The butterfly fauna of the Sacramento Valley, California. J. Res. Lepidoptera 13:73-82,115-122,137-148.
-----. 1978. A new weedy host for the buckeye, Precis coenia (Nymphalidae). J. Lepidoptera Soc. 32:224.
SINGER, M. C. 1971. Evolution of food-plant preference in the butterfly Euphydryas editha. Evolution 25:383-389.
SINGER, M. C., D. NG, AND C. D. THOMAS. 1988. Heritability of oviposition preference and its relationship to offspring performance within a single insect population. Evolution 42:977-985.
SINGER, M. C., C. D. THOMAS, H. L. BILLINGTON, AND C. PARMESAN. 1989. Variation among conspecific insect populations in the mechanistic basis of diet breadth. Anita. Behar. 37:751-759.
STAMP, N. E. 1992. Susceptibility of specialist versus generalist caterpillars to invertebrate predators. Oecologia (Berlin) 92: 124-129.
STRONG, D. R., J. H. LAWTON, AND S. R. S. SOUTHWOOD. 1984. Insects on plants: community patterns and mechanisms. Harvard Univ. Press, Cambridge, MA.
THOMAS, C. D., D. NG, M. C. SINGER, J. L. B. MALLET, C. PARMESAN, AND H. L. BILLINGTON. 1987. Incorporation of a European weed into the diet of a North American herbivore. Evolution 41:892-901.
THOMPSON, J. N., AND O. PELLMYR. 1991. Evolution of oviposition behavior and host preference in the Lepidoptera. Annu. Rev. Entomol. 36:65-89.
VIA, S. 1984. The quantitative genetics of polyphagy in an insect herbivore. I. Genotype environment interaction in larval performance on different host species. Evolution 38:881-895.
-----. 1986. Genetic covariance between oviposition preference and larval performance in an insect herbivore. Evolution 40: 778-785.
-----. 1990. Ecological genetics and host adaptation in herbivorous insects: the experimental study of evolution in natural and agricultural systems. Annu. Rev. Entomol. 35:421-446.
|Printer friendly Cite/link Email Feedback|
|Author:||Camara, Mark D.|
|Date:||Jun 1, 1997|
|Previous Article:||Allozymic differentiation in response to laboratory demographic selection of Drosophila melanogaster.|
|Next Article:||Intraspecific phylogeography of North American highland fishes: a test of the Pleistocene vicariance hypothesis.|